The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation techniques, or persistence mechanisms were found. The skill defines a methodical workflow for implementation planning including task decomposition and internal reviews.
[COMMAND_EXECUTION]: The skill instructions guide the agent to include shell commands like pytest and git commit within its generated output. These are provided as template steps for a developer or agent to follow during the implementation phase and do not represent immediate or hidden command execution by the skill itself.
[DATA_EXFILTRATION]: The skill ingests project specifications and requirements to generate documentation. No instructions were found that would cause the agent to exfiltrate this sensitive project data to external or untrusted network locations.
[PROMPT_INJECTION]: The skill incorporates a specialized review process using a subagent (guided by plan-document-reviewer-prompt.md) to verify that implementation plans are complete and align with the provided specifications. This structured review loop acts as a safeguard against errors or deviations in the agent's output.

writing-plans